1. Field of the Invention
The present invention relates to a heterodyne coherent type position detection apparatus, and, more particularly, to an apparatus suitably used for accurate alignment in an apparatus for manufacturing semiconductors.
2. Related Background Art
In recent years, a projection type exposing apparatus, so-called a stepper has been widely used as an apparatus for transcribing fine patterns on the surface of a semiconductor devices or the like onto the wafer of a semiconductor. In particular, since high density mount LSIs manufactured by the above-described technology have been desired recently, finer pattern must be transcribed to wafers. In order to achieve this, more accurate alignment is necessary.
An accurate position detection apparatus in accordance with the heterodyne coherent method is known, as disclosed in, for example, Japanese Patent Laid-Open No. 62-261003 and has been known U.S. Pat. No. 4,710,026 has been known.
The former apparatus is arranged in such a manner that its aligning light source comprises a Zeeman laser utilizing a Zeeman effect which emits luminous flux including p-polarized light and s-polarized light having slightly different frequencies. The luminous flux emitted from the Zeeman laser is, by a polarizing beam splitter, divided into two fluxes: p-polarized light having a frequency of f.sub.2 and s-polarized light having a frequency of f.sub.2. The thus formed two fluxes irradiate a diffraction grating mark formed on a reticle (or a mask) from predetermined two directions via two corresponding reflecting mirrors. Since the mask has an opening neighboring the diffraction grating mark, a portion of each of the two fluxes can, from predetermined two directions, irradiate the diffraction grating mark formed on the wafer after passing through the opening.
As a result, the diffraction grating mark on the mask and that on the wafer respectively generate diffracted beams. The diffracted beams (p-polarized light and s-polarized light) emitted from the mask mark interfere with each other via a polarizing plate. As a result, a single optical beat signal is generated. Also the diffracted beams from the wafer mark interfere with each other via the polarizing plate. As a result, another optical beat signal is generated. The thus generated two optical beat signals are photoelectrically detected so that the relative phase difference between the two signals is detected. Since the phase difference corresponds to the quantity of deviation between the two luminous fluxes intersecting on the diffraction grating mark and the substrate, an accurate alignment can be achieved by relatively moving the reticle and the wafer in such a manner that the phase difference becomes zero.
However, the above described conventional apparatuses have a problem in that an error takes place due to the following factors:
The first factor is that the perfect splitting of polarized beams cannot be achieved if a slight positional deviation from the designed positions takes place between the active plane of a polarization beam splitter of a polarization splitting device such as the above-described polarization beam splitter for splitting two polarized beams from the Zeeman light source and the polarizing plane on which the two polarized beams are made incident. As a result, the split luminous flux includes not only the main polarized beam component but also a small quantity the other polarized beam component, that causes noise.
The above-described positional deviation is caused by errors that have mainly taken place at the time of manufacturing the semiconductors. However, the polarization splitting device such as the polarization beam splitter cannot perfectly split p-polarized light and s-polarized light. Therefore, even if the optical system comprising the above-described devices can be mounted without the manufacturing errors, each of polarized beams which has been split includes a small quantity of noise which is different in the polarization component and in the frequency from the main polarized beam component. The split polarized beams become coherent to each other so that an optical beat is generated between the main polarized beam component and the noise component, causing an excessive error to appear in the optical beat signal which has been photoelectrically detected.
A second factor is caused by the fact that linearly polarized luminous flux is, strictly, deformed into elliptically polarized light due to the passage of the flux through optical devices such as a polarization beam splitter and a reflecting mirror. Usually, a laser beam emitted from a light source reaches the diffraction grating mark on the mask or on the reticle after it has been subjected to one or more reflections. During the process of the passage of the laser beam, a noise factor of the type shown in FIGS. 4 and 7 is generated. That is, noise component Pr (frequency: f.sub.1) is generated in direction x which is perpendicular to main polarized beam component Pry (frequency: f.sub.1) in direction y. Furthermore, noise component Ny (frequency: f.sub.2) is generated in direction y which is perpendicular to noise component Nx (frequency: f.sub.2) in direction x. Although the noise components (Prx and Nx) which are generated perpendicularly to the main polarizing direction can be eliminated, the noise component Ny in the same direction as main polarized beam component Pry cannot be eliminated. Therefore, beats generate between the main polarized beam component Pry (frequency: f.sub.1) and the noise component Ny (frequency: f.sub.2) in the same direction as the main polarized beam component Pry, causing the optical beam signal whose position is to be detected to include an excessive error.
Now, the above-described problem will be analyzed with reference to the two diffracted beams to be photoelectrically detected.
The amplitude D.sub.1 of the diffracted light consisting of the main polarized beam component having the frequency of f.sub.1 and the noise component having the frequency of f.sub.2 included in the main polarized beam component can be expressed as follows: EQU D.sub.1 =U.sub.1 exp[-i(k.sub.1 t-.phi..sub.1)]+V.sub.2 exp[-i(k.sub.2 t-.phi..sub.2)] (1)
On the other hand, the amplitude D.sub.2 of the diffracted light consisting of the main polarized beam component having the frequency of f.sub.2 and the noise component having the frequency of f.sub.1 included in the main polarized beam component can be expressed as follows: EQU D.sub.2 =U.sub.2 exp[-i(k.sub.2 t+.phi..sub.2)]+V.sub.1 exp[-i(k.sub.1 t+.phi..sub.1)] (2)
where U.sub.1 and V.sub.2 respectively represent the amplitude of the main polarized beam component (frequency: f.sub.1) and that of the noise component (frequency: f.sub.2) of the diffracted light having the amplitude D.sub.1, U.sub.2 and V.sub.1 respectively represent the amplitude of the main polarized beam component (frequency: f.sub.2) and that of the noise component (frequency: f.sub.1) of the diffracted light having the amplitude D.sub.2, k.sub.1 (=2.pi.f.sub.1) and k.sub.2 (=2.pi.f.sub.1) represent the wave numbers, and .phi..sub.1 and .phi..sub.2 represent the phase changes generated when the diffracted beams which respectively have the frequencies f.sub.1 and f.sub.2 are subjected to the diffraction by means of the diffraction grating.
Intensity of the optical beam signal which is photoelectrically detected by a detector can be expressed as follows from thus obtained Equations (1) and (2): ##EQU1## Expansion of Equation (3) gives ##EQU2##
The optical beat signal obtained from Equation (4) originally includes only the phase component .phi..sub.1 +.phi..sub.2 representing the positional information of the diffraction grating. It is apparent that the optical beat signal further includes other phase components (.phi..sub.1 -.phi..sub.2) and (-.phi..sub.1 +.phi..sub.2). Assuming that V.sub.1 and V.sub.2 are sufficiently small with respect to U.sub.1 and U.sub.2, the phase error e can be expressed by the following equation: ##EQU3##
Assuming that the light intensity ratio of the polarized components perpendicular to each other and included in each of two luminous fluxes split by the above-described polarization beam splitter is 1:1000 (.vertline.V.sub.2 .vertline..sup.2 :.vertline.U.sub.1 .vertline..sup.2 or .vertline.U.sub.2 .vertline..sup.2 :.vertline.V.sub.1 .vertline..sup.2) and U.sub.1 =U.sub.2 and V.sub.1 =V.sub.2, the phase error e according to Equation (5) becomes ##EQU4##
Therefore, the optical beat signal which is photoelectrically detected inevitably includes an error which cannot be disregarded.
The conventional apparatus has been arranged in such a manner that the pitch of interference fringes generated when the mark on the diffraction grating formed on the substrate is illuminated with two luminous fluxes having different frequencies from two directions becomes half of the pitch of the diffraction gratings. Therefore, the phase difference between the optical beat signal including positional information of the diffraction grating mark and the reference optical beat signal changes by 2.pi. whenever the relative deviation between the two luminous fluxes and the substrate becomes the half pitch. That is, the range which can be detected as the phase difference becomes 2.pi..
Then, assuming that the pitch of the mark on the diffraction grating formed on the substrate is P, the above-described phase error e can be converted into the following relative deviation .delta. between the two luminous fluxes and the substrate: ##EQU5##
Assuming that the pitch P of the mark on the diffraction grating is 10 .mu.m, the above-described phase error of 3.6.degree. can be converted into the deviation .delta. expressed by: ##EQU6## It is apparent that the above-described error cannot be disregarded.
When, for example, a 4-megabit VLSI is manufactured, a printing line width which is about 0.6 to 0.7 .mu.m is necessary. In order to achieve this, the detector must have an accuracy of at least 10% (which is 0.06 to 0.07 .mu.m) of the line width (about 0.6 to 0.7 .mu.m). However, the error obtained in accordance with Equation (8) causes a difficulty in achieving an alignment at the time of subjecting 4-megabit VLSI to an exposure.
The above-described conventional apparatus has experienced another detection accuracy problem due to diffracted light from a field stop. In order to overcome the problem of this type, it is preferable that the field stop be positioned in conjugation with the position of a wafer in an optical system for transmitting alignment light when the above-described heterodyne type position detection is conducted. Thus, the illumination region of the mark on the diffraction grating of the wafer may be restricted. The reason for this is that the detector should be protected from reflected light acting as noise and generated when a portion of alignment light is applied to the pattern in the transcription region or another alignment mark.